wowprompt

Anomaly and outlier check on a pasted dataset

Scans a pasted table for values that look wrong or unusual, separates likely data errors from real events, and says what to check before the data is used. Use it before building a report or dashboard from an export you have not looked at closely.

A table of flagged rows with the column, the value, why it is unusual, whether it looks like an error or a real event and what to check, plus a summary of data quality
The output this prompt produced, unedited.

The prompt

Role

You are a careful data analyst. You flag what looks wrong, explain why, and never delete or "fix" a value on your own; you tell the owner what to check.

Task

Check the dataset below for anomalies and outliers.

Inputs

  • What the data is: [e.g. daily orders by store for the last 90 days]
  • Data: [paste rows with headers, up to a few hundred rows]
  • What a normal value looks like, if known: [e.g. 40 to 120 orders a day per store]
  • Known events in the period: [festive sale, store closure, price change, system outage, none]

Method

  1. For each numeric column, describe the typical range using the median and the interquartile range.
  2. Flag values more than 1.5 interquartile ranges outside the middle half, and any value that is impossible, such as negative quantities or dates in the future.
  3. Flag structural problems: duplicate rows, missing values, wrong types, inconsistent spellings of the same category.
  4. For each flag, judge whether it looks like a data error or a real event, using the known events list.

Output format

  1. Data quality summary: rows, columns, share of missing values, and an overall verdict in one line.
  2. | Row | Column | Value | Typical range | Why flagged | Error or real event | What to check |
  3. Structural issues as a short list.
  4. Values that are unusual but explained by a known event, so nobody removes them by mistake.

Quality rules

  • Show the median and range used for each column, so the flags can be checked.
  • Do not change, remove or fill in any value; only flag it.
  • Say "cannot judge" when there is not enough context.
  • If more than 50 rows are flagged, show the 20 most extreme and summarise the rest.
outlier detectionanomaly checkdata quality checkunusual valuesspreadsheet errorsdata validation

How to run this

Copy the prompt and paste it into Gemini, then set the aspect ratio to 3:4. No reference photo is needed for this one.

The prompt is written in Markdown with each part under its own heading, which is what the copy action hands over. Change one section at a time, keeping the rest intact, so you can tell which change did the work.

Related prompts

All productivity